Research map / Adversarial robustness and security / Attacks on text and graphs

Boosting the Adversarial Robustness of Graph Neural Networks: An OOD Perspective

ICLR 2024 · Adversarial robustness and security › Attacks on text and graphs · approach: Graph neural network robustness

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Studies adversarial training for graph neural networks from an out-of-distribution perspective to improve robustness against structural graph attacks without relying on hand-designed edge properties.

Problem

Existing graph attack defenses rely on properties for identifying adversarial edges, making them vulnerable to adaptive white-box attacks; adversarial training for GNNs is a candidate but the abstract is cut off before its limitations are stated.

Approach

Frames adversarial perturbations of graph structure as out-of-distribution examples to guide adversarial training of GNNs, avoiding reliance on artificially designed edge properties. The abstract is truncated, so the specific mechanism is not stated.

Closest papers by meaning

• Are Defenses for Graph Neural Networks Robust? (NeurIPS 2022 · similarity 0.94)
• Adversarial Training for Graph Neural Networks: Pitfalls, Solutions, and New Directions (NeurIPS 2023 · similarity 0.92)
• Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks (AAAI 2024 · similarity 0.92)
• If You Want to Be Robust, Be Wary of Initialization (NeurIPS 2024 · similarity 0.91)
• Bounding the Expected Robustness of Graph Neural Networks Subject to Node Feature Attacks (ICLR 2024 · similarity 0.91)
• Robust Graph Neural Networks via Unbiased Aggregation (NeurIPS 2024 · similarity 0.89)